Fantastic Animals and Where to Find Them - Segment Any Marine Animal with Dual SAM

Aug 25, 2026

Uploaded by Theodora

Marine Animal Segmentation (MAS) focuses on accurately delineating animals in underwater scenes where lighting varies, turbidity distorts colors, and both camera and targets move. Prior approaches often miss long-range contextual cues and fail to model connectivity among discrete pixels. This CVPR work introduces Dual-SAM, a dual-structure extension of Segment Anything Model that improves marine feature learning, adds Multi-level Coupled Prompts, and fuses multi-level encoder representations. A Criss-Cross Connectivity Prediction with mutual supervision yields state-of-the-art results on five MAS datasets.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Abstract

Introduction

This CVPR paper is the Open Access version, provided by the Computer Vision Foundation. Except for this watermark, it is identical to the accepted version; the final published version of the proceedings is available on IEEE Xplore. Fantastic Animals and Where to Find Them: Segment Any Marine Animal with Dual SAM Pingping Zhang* Tianyu Yan Yang Liu Huchuan Lu School of Future Technology, School of Artificial Intelligence, Dalian University of Technology, China [2981431354@mail.dlut.edu.cn](2981431354@mail.dlut.edu.cn) ; {zhpp,ly,[lhchuan](lhchuan}@dlut.edu.cn)[}](lhchuan}@dlut.edu.cn)[@dlut.edu.cn](lhchuan}@dlut.edu.cn) Abstract As an important pillar of underwater intelligence, Marine Animal Segmentation (MAS) involves segmenting animals within marine environments. Previous methods don’t excel in extracting long-range contextual features and overlook the connectivity between discrete pixels. Recently, Segment Anything Model (SAM) offers a universal framework for general segmentation tasks. Unfortunately, trained with natural images, SAM does not obtain the prior knowledge from marine images. In addition, the single-position prompt of SAM is very insufficient for prior guidance. To address these issues, we propose a novel feature learning framework, named Dual-SAM for high-performance MAS. To this end, we first introduce a dual structure with SAM’s paradigm to enhance feature learning of marine images. Then, we propose a Multi-level Coupled Prompt (MCP) strategy to instruct comprehensive underwater prior information, and enhance the multi-level features of SAM’s encoder with adapters. Subsequently, we design a Dilated Fusion Attention Module (DFAM) to progressively integrate multi-level features from SAM’s encoder. Finally, instead of directly predicting the masks of marine animals, we propose a Criss-Cross Connectivity Prediction (C 3P) paradigm to capture the inter-connectivity between discrete pixels. With dual decoders, it generates pseudo-labels and achieves mutual supervision for complementary feature representations, resulting in considerable improvements over previous techniques. Extensive experiments verify that our proposed method achieves state-of-the-art performances on five widely-used MAS datasets. The code is available at [https://github.com/Drchip61/Dual](https://github.com/Drchip61/Dual SAM)[ ](https://github.com/Drchip61/Dual SAM)[SAM](https://github.com/Drchip61/Dual SAM). 1. Introduction Underwater ecosystems contain a wide variety of marine life, from microscopic plankton to colossal whales. These ecosystems are crucial roles for the earth’s environmental *Corresponding author | ViT Encoder |

Mask Decoder
| | --- | --- | | | | | Mask Prompt | | (a) SAM’s Prompt | Mask Prompt |

| ViT Encoder | | --- | --- | --- | (b) Our Multi-level Prompt (c) Mutual Supervision (d) High Performance of Dual-SAM Figure 1 . Our inspirations and advantages. (a) Single-position prompt of SAM. (b) Our multi-level prompt. (c) Mutual supervision for our Dual-SAM’s decoders. (d) Our Dual-SAM delivers high performances on multiple datasets. balance. Accurate and efficient Marine Animal Segmentation (MAS) is vital for understanding species’ distributions, behaviors, and interactions within the submerged world. However, unlike conventional terrestrial images, underwater images include variable lighting conditions, water turbidity, color distortion, and the movement of both cameras and subjects. Traditional segmentation techniques, developed primarily for terrestrial settings, often fall short when applied to the underwater domain. Consequently, methods designed to tackle the unique aspects of the marine environment are urgently required for underwater intelligence. With the advent of deep learning, Convolutional Neural Networks (CNNs) [15, 20] lead to a new era for image segmentation. In fact, CNNs demonstrate a remarkable ability to extract intricate features, which mak

FAQs

What problems does Dual-SAM target in marine animal segmentation?

Dual-SAM addresses the lack of long-range contextual feature extraction, insufficient connectivity modeling between discrete pixels, and the mismatch between SAM’s natural-image training and marine-image priors.

How does Dual-SAM improve prompts and feature learning for underwater images?

It introduces a dual structure aligned with SAM’s paradigm and enhances encoder multi-level features with adapters. It also uses a Multi-level Coupled Prompt (MCP) strategy to provide comprehensive underwater prior information.

What is the Criss-Cross Connectivity Prediction (C3P) paradigm used for?

Instead of directly predicting masks, C3P captures inter-connectivity between discrete pixels using dual decoders that generate pseudo-labels and enable mutual supervision for complementary representations.

Aug 25, 2026

Uploaded by Theodora